14results about How to "Reduce bandwidth consumption" patented technology

The invention discloses a high-concurrency trusted computing remote certification method and system based on equipment type adaptation, and relates to the technical scheme that the method comprises the following steps: receiving a registration request which is sent by a client and contains an equipment type identifier; creating a thread pool, packaging the received registration request into a registration processing task, scheduling the registration processing task, analyzing an equipment type field in the registration request, and storing a mapping relationship between a client identifier and an equipment type to a local configuration file; in the challenge certification stage, querying the equipment type of the client from the local configuration file, and selecting an adaptive remote certification protocol according to a preset mapping rule; and packaging response verification operation of the server to the client into a proof processing task, and submitting the proof processing task to a thread pool for high-concurrency scheduling. The method can adapt to multi-form client requests, and can efficiently process large-scale concurrent remote attestation requests at the same time.

This application provides a method and system for managing large file caches automatically in a browser, relating to the field of cache management technology. This application intercepts resource requests and generates resource descriptors during the loading of a 3D geographic information scene. Based on browser type, resource type, network caching policy, and storage quota, it determines the automatic caching threshold and target resource data volume to identify resources exceeding the threshold. A scene dependency pool containing necessary resources for the first screen and expected resources for adjacent view areas is generated according to scene dependency identifiers. Resources exceeding the threshold are fragmented and written to the browser database to form shadow cache records. By combining address keys, version verification information, resource-level integrity bitmaps, and fragment-level integrity bitmaps, it performs hit reconstruction, missing data filling, changed fragment replacement, and pool-level eviction. This improves the caching recoverability of large file resources on the browser side, the stability of first-screen loading, version update efficiency, and capacity management accuracy.

The application provides a kind of transmission layer data packet management method based on fragmentation and reorganization, belong to computer network technical field.Each device includes a server and multiple clients;The process of the first device transmitting messages to the second device is as follows: multiple clients of the first device transmit messages to a sending queue;The server of the first device monitors the sending queue, fragments the message data packet to obtain multiple sub-packets;The server of the second device receives the sub-packets sent by the server of the first device, reassembles the sub-packets, and sends them to a receiving queue;The clients of the second device block and wait for messages from the receiving queue, filter messages by analyzing the identity field in the message header, and process data packets that match their own identity.This method dynamically adjusts the packet size based on the network environment state, avoiding the packet loss or truncation problem caused by MTU limitation on high latency links in the traditional fixed packet mode.

This disclosure relates to the field of semiconductor technology, providing a semiconductor device to address the technical problem that high-bandwidth memory technology still has significant room for improvement. The semiconductor device includes a chip stack structure, comprising: a first semiconductor chip and a second semiconductor chip stack structure stacked along a first direction; the first semiconductor chip includes a controller, a computing module, and an interface module; the controller is configured to: perform a first access operation on stored data in the second semiconductor chip stack structure and send the stored data to the interface module to allow external devices to access the stored data in the second semiconductor chip stack structure through the interface module; and perform a second access operation on the stored data in the second semiconductor chip stack structure and send the stored data to the computing module to allow the computing module to perform computational operations on the stored data. This improves integration and shortens data transmission paths, reducing bandwidth consumption and energy loss.

The invention discloses an auto-regression and mask diffusion double-normal-form fusion language model hardware acceleration system based on an FPGA (Field Programmable Gate Array), and relates to the field of FPGA and machine learning. The invention provides a model core calculation process-oriented hardware acceleration architecture by aiming at a language model fusing double normal forms of an autoregression model and a mask diffusion model and adopting the characteristics of revising an attention mechanism and parallel generation of a KV cache. The method disclosed by the invention is implemented by taking an Eso-LMs (Esolic Language Models) model as an example. The system comprises a layer normalization and adaptive layer normalization modulation module, a QKV projection module, a rotation position coding application module, a KV cache management module, a multi-head attention calculation module, an output projection and residual connection module and a multi-layer perceptron module. By optimizing the calculation sequence and the data flow of each module in the model calculation process and adopting the pipeline parallel and resource reuse technology, the efficient reasoning acceleration of the language model fusing the autoregression and mask diffusion double normal forms on the FPGA platform is realized, and the model reasoning speed is obviously improved. According to the system, the advantage of FPGA customizable hardware acceleration is fully exerted, the utilization efficiency of hardware resources is improved, data pipeline blockage is eliminated, calculation delay and storage overhead are reduced by reconstructing the data flow direction and constructing a whole-process pipeline processing architecture, and the system is suitable for the deployment requirement of an edge calculation scene.

This invention proposes a wearable device interaction method and apparatus, and a computer device based on memory information. The method includes: collecting multimodal raw data through a wearable device; performing event recognition and semantic extraction on the raw data of each modality, and outputting a single-modal semantic description; comprehensively analyzing all single-modal semantic descriptions within the same time period to generate an event chain and event context description; saving an event table, an association information table, and an index table to a database, wherein the event table stores the event chain and event context description, the association information table stores the relationships between events, and the index table establishes a data retrieval channel; automatically generating reminder tasks based on the data in the event table, association information table, and index table, and pushing them to the user when the task triggering conditions are met, wherein the task triggering conditions include at least one of time, location, scenario, and unfinished tasks. This invention achieves efficient and accurate active and passive intelligent interaction, improving user experience and system efficiency.

This invention discloses a personnel behavior recognition and early warning system for specific areas of a power plant, relating to the field of intelligent industrial safety monitoring technology. It includes a data acquisition module, an edge computing module, and a central processing platform. The data acquisition module is responsible for collecting and initially processing multi-source heterogeneous data, including personnel identity, behavior, location, and environmental parameters. The edge computing module performs lightweight processing and localized decision-making on the raw data, reducing the load on the central processing platform. The central processing platform achieves full-process safety control of high-risk areas through multimodal data fusion, dynamic access control, and trajectory analysis. In this invention, lightweight AI models are embedded in various cameras of the employee badge authentication terminal and visual perception unit, enabling preliminary inference to be performed directly on the employee badge authentication terminal and camera, thus equipping the employee badge authentication terminal and camera with edge computing capabilities. Independent edge servers or industrial control computers are deployed near the data sources in various areas of the power plant as regional edge computing nodes.

This invention discloses an adaptive data acquisition, signing, and encapsulation method and system for the Internet of Things (IoT). The method includes: dynamically determining a batch signing threshold based on environmental parameters; acquiring data in real time, calculating and caching hash values, and encapsulating the data and its corresponding hash values ​​into data packets for transmission to the cloud; when the number of transmitted data packets reaches the batch signing threshold, triggering a signing operation, generating an overall digital signature for all cached hash values ​​using a built-in private key, and encapsulating the set of all hash values ​​and the overall digital signature into a signature packet for transmission to the cloud for cloud verification of data integrity and batch confirmation of data packet authenticity. This invention changes the traditional strongly bound encapsulation mode of "data + signature," separating the data packets and signature packets into separate encapsulations, changing the traditional "signature for every packet" to "acquisition and transmission as needed, batch signing," greatly reducing the frequency of asymmetric encryption by hardware security chips, eliminating transmission congestion caused by encryption, and significantly increasing system throughput.

The invention discloses a multi-target detection method based on task driving, and the method comprises the steps: S1, building a data set, and carrying out the preprocessing: collecting image data of a to-be-detected and monitored region according to a target detection task, carrying out the marking of the obtained image data, making a data set, and carrying out the preprocessing of the data set; s2, constructing and training a multi-target detection model: constructing a multi-target detection network, and training the model by using a data set; S3, detecting and verifying: locally deploying the multi-target detection model, accessing multiple paths of real-time video streams into the multi-target detection model for multi-target detection, and judging whether a target exists in a monitoring area or not; and comparing and analyzing a multi-channel video stream pushing result with a polling algorithm pushing result, and outputting a final result. According to the method, a UIB module lightweight algorithm and a polling algorithm are combined to save bandwidth consumption, and a PIOUv2 loss function is combined to improve detection precision, accuracy, real-time performance and adaptability under the condition of limited hardware and limited bandwidth.

The invention discloses an enterprise-based data outbound system, which belongs to the technical field of data transmission, and comprises a data outbound capability layer, and the data outbound capability layer comprises the following modules: a compliance auditing module, which is used for carrying out compliance check on cross-border data flow to ensure that data transmission meets regulatory requirements of a target country or region, the security guarantee module is used for guaranteeing the security and transparency of the data in the cross-border transmission process, and the data flow optimization module is used for optimizing the data flow based on real-time network conditions, regulation requirements and data transmission efficiency factors. According to the invention, automatic compliance review and multiple security protection of cross-border data flow are realized, and the risk of violation of an enterprise is remarkably reduced. Through intelligent routing optimization and edge calculation, the transmission efficiency is effectively improved, and delay and bandwidth consumption are reduced. The architecture can be seamlessly integrated with an existing enterprise framework, has good compatibility and expansibility, and supports an enterprise to realize safe, compliant and efficient data flow management in a global range.

This application relates to a method and system for comparing table data in a row-oriented storage database. The method includes: acquiring the database to be verified; collecting business characteristics, constructing a B+ tree index, and partitioning the database; sequentially locating the difference partitions, difference rows, and difference fields based on the B+ tree index and partitioning combined with hash calculation; and outputting the comparison results based on the difference fields. This embodiment achieves rapid location and on-demand reading of difference partitions in a row-oriented storage database. By sequentially locating the difference partitions, difference rows, and difference fields based on the B+ tree index and partitioning combined with hash calculation, it achieves rapid identification of the partitions containing the difference data, avoiding a full scan of irrelevant data, thereby significantly reducing disk I / O, memory usage, and bandwidth consumption, improving verification efficiency, and reducing resource overhead.

The invention provides a convolutional neural network inference method based on a core grain architecture, and the method comprises the steps: S1, obtaining to-be-processed data and a pre-trained convolutional neural network, and S2, distributing a mapping region for each network layer in the core grain architecture according to a graph division strategy; s3, segmenting the feature map and the weight tensor corresponding to each network layer into a plurality of feature blocks and a plurality of weight blocks, and sequentially distributing the plurality of weight blocks corresponding to each network layer to a plurality of calculation units in the corresponding mapping area, determining a loading sequence of transmitting each feature block corresponding to each network layer to different computing units in the corresponding mapping region to obtain an initial mapping scheme of the convolutional neural network; s4, performing multi-round iteration on the initial mapping scheme to obtain a target mapping scheme; and S5, based on the target mapping scheme, loading the convolutional neural network to a calculation unit corresponding to the core particle architecture to realize reasoning of the to-be-processed data.

The invention provides a data anomaly adaptive detection intelligent early warning method based on edge calculation. The method comprises the following steps: S1, acquiring monitoring data; s2, performing regression analysis, and determining tendency; s3, if the trend exists, judging abnormity based on the LSTM; s4, if the data does not have the tendency, judging whether the data accords with bimodal distribution based on K-means clustering, and if the data accords with the bimodal distribution, judging that the data is abnormal based on a K-means clustering result; s5, if no tendency exists and no bimodal distribution exists, judging abnormity based on LSTM and box diagram analysis; s6, correcting a normal value or retaining a preliminary abnormal value; s7, judging the abnormal value as a real abnormal value; and S8, outputting an equipment fault early warning signal or a normal monitoring signal according to a result of the S7. According to the method, the limitation that only known data change rules can be processed in the prior art is overcome, the adaptability to variable environments is enhanced, and a solid technical support is provided for stable operation and timely fault early warning of Internet of Things equipment.

This application relates to the field of data processing technology, and discloses an image processing method, apparatus, device, medium, and product. The method includes: acquiring a first interface image to be processed and a second interface image preceding the first interface image, when dragging does not display window content; dividing the first interface image and the second interface image into multiple image blocks; identifying the first image block in the first interface image that has changed; when a window dragging operation exists, encoding the first image block at the corresponding position according to the second image block in the second interface image to generate encoded image blocks; and sending the encoded image blocks corresponding to each first image block to a client. This application utilizes the characteristic that most pixels are the same between the first image block and the second image block in the previous second interface image, so that the generated encoded image block has a smaller data volume than the first image block, which can reduce bandwidth consumption in window dragging scenarios.